Hey David, thanks for your quick reply!
Yes I need GPS data for each frame, because we need to localize the objects found on these images as precisely as possible. I am able to read detailed GPS data from the MP4 file, i.e. with more than one data point per second.
Now I was thinking of writing a (Python) script that would first calculate the time of a frame within the video and write that (start + elapsed time) as original date exif information into the JPG file. Then as a second step I would read in all GPS data points from the GPX file, go through each video frame, get its time stamp in order to find the GPS data point that is closest to that time stamp. Or even better: Get the GPS data point that is before that image time stamp and the one after and then interpolate these GPS data points.
Would this be a good approach? And isn’t there any tool that achieves this already?